Search

Traditional computer vision models often necessitate extensive data acquisition, annotation, and validation. These models frequently struggle in real-world applications, resulting in high false positive and negative rates, and exhibit poor adaptability to new scenarios, often requiring costly retraining. To address these issues, we present Ethosight, a flexible and adaptable zero-shot video analytics system. Ethosight begins from a clean slate based on user-defined video analytics, specified through natural language or keywords, and leverages joint embedding models and reasoning mechanisms informed by ontologies such as WordNet and ConceptNet. Ethosight operates effectively on low-cost edge devices and supports enhanced runtime adaptation, thereby offering a new approach to continuous learning without catastrophic forgetting. We provide empirical validation of Ethosight's promising effectiveness across diverse and complex use cases, while highlighting areas for further improvement. A significant contribution of this work is the release of all source code and datasets to enable full reproducibility and to foster further innovation in both the research and commercial domains.

We consider two jointly stationary and ergodic random measures $\xi$ and $\eta$ on $\mathbb{R}^d$ with equal finite intensities, assuming $\xi$ to be diffuse. An allocation is a random mapping taking $\mathbb{R}^d$ to $\mathbb{R}^d\cup\{\infty\}$ in a translation invariant way. We construct allocations transporting the diffuse $\xi$ to arbitrary $\eta$, under the mild condition of existence of an `auxiliary' point process which is needed only in the case when $\eta$ is diffuse. When that condition does not hold we show by a counterexample that an allocation transporting $\xi$ to $\eta$ need not exist.

A cognitive architecture aimed at cumulative learning must provide the necessary information and control structures to allow agents to learn incrementally and autonomously from their experience. This involves managing an agent's goals as well as continuously relating sensory information to these in its perception-cognition information stack. The more varied the environment of a learning agent is, the more general and flexible must be these mechanisms to handle a wider variety of relevant patterns, tasks, and goal structures. While many researchers agree that information at different levels of abstraction likely differs in its makeup and structure and processing mechanisms, agreement on the particulars of such differences is not generally shared in the research community. A binary processing architecture (often referred to as System-1 and System-2) has been proposed as a model of cognitive processing for low- and high-level information, respectively. We posit that cognition is not binary in this way and that knowledge at any level of abstraction involves what we refer to as neurosymbolic information, meaning that data at both high and low levels must contain both symbolic and subsymbolic information. Further, we argue that the main differentiating factor between the processing of high and low levels of data abstraction can be largely attributed to the nature of the involved attention mechanisms. We describe the key arguments behind this view and review relevant evidence from the literature.

This paper introduces a new metamodel-based knowledge representation that significantly improves autonomous learning and adaptation. While interest in hybrid machine learning / symbolic AI systems leveraging, for example, reasoning and knowledge graphs, is gaining popularity, we find there remains a need for both a clear definition of knowledge and a metamodel to guide the creation and manipulation of knowledge. Some of the benefits of the metamodel we introduce in this paper include a solution to the symbol grounding problem, cumulative learning, and federated learning. We have applied the metamodel to problems ranging from time series analysis, computer vision, and natural language understanding and have found that the metamodel enables a wide variety of learning mechanisms ranging from machine learning, to graph network analysis and learning by reasoning engines to interoperate in a highly synergistic way. Our metamodel-based projects have consistently exhibited unprecedented accuracy, performance, and ability to generalize. This paper is inspired by the state-of-the-art approaches to AGI, recent AGI-aspiring work, the granular computing community, as well as Alfred Korzybski's general semantics. One surprising consequence of the metamodel is that it not only enables a new level of autonomous learning and optimal functioning for machine intelligences, but may also shed light on a path to better understanding how to improve human cognition., Comment: arXiv admin note: text overlap with arXiv:2008.12879

We consider two jointly stationary and ergodic random measures $\xi$ and $\eta$ on the real line $\mathbb{R}$ with equal intensities. An allocation is an equivariant random mapping from $\mathbb{R}$ to $\mathbb{R}$. We give sufficient and partially necessary conditions for the existence of allocations transporting $\xi$ to $\eta$. An important ingredient of our approach is to introduce a transport kernel balancing $\xi$ and $\eta$, provided these random measures are mutually singular. In the second part of the paper, we apply this result to the path decomposition of a two-sided Brownian motion into three independent pieces: a time reversed Brownian motion on $(-\infty,0]$, an excursion distributed according to a conditional It\^o's law and a Brownian motion starting after this excursion. An analogous result holds for Bismut's excursion law., Comment: 22 pages, 2 figures. This paper is published by https://projecteuclid.org/euclid.aihp/1539849799

The concept of "task" is at the core of artificial intelligence (AI): Tasks are used for training and evaluating AI systems, which are built in order to perform and automatize tasks we deem useful. In other fields of engineering theoretical foundations allow thorough evaluation of designs by methodical manipulation of well understood parameters with a known role and importance; this allows an aeronautics engineer, for instance, to systematically assess the effects of wind speed on an airplane's performance and stability. No framework exists in AI that allows this kind of methodical manipulation: Performance results on the few tasks in current use (cf. board games, question-answering) cannot be easily compared, however similar or different. The issue is even more acute with respect to artificial *general* intelligence systems, which must handle unanticipated tasks whose specifics cannot be known beforehand. A *task theory* would enable addressing tasks at the *class* level, bypassing their specifics, providing the appropriate formalization and classification of tasks, environments, and their parameters, resulting in more rigorous ways of measuring, comparing, and evaluating intelligent behavior. Even modest improvements in this direction would surpass the current ad-hoc nature of machine learning and AI evaluation. Here we discuss the main elements of the argument for a task theory and present an outline of what it might look like for physical tasks., Comment: accepted to the Ninth Conference on Artificial General Intelligence (AGI-16)

According to the Dudley-Wichura extension of the Skorohod representation theorem, convergence in distribution to a limit in a separable set is equivalent to the existence of a coupling with elements converging a.s. in the metric. A density analogue of this theorem says that a sequence of probability densities on a general measurable space has a probability density as a pointwise lower limit if and only if there exists a coupling with elements converging a.s. in the discrete metric. In this paper the discrete-metric theorem is extended to stochastic processes considered in a widening time window. The extension is then used to prove the separability version of the Skorohod representation theorem.

If the inter-arrival time distribution of a renewal process is regularly varying with index $\alpha\in\left( 0,1\right) $ (i.e. the inter-arrival times have infinite mean) and if $A\left( t\right) $ is the associated age process at time $t$. Then we show that if $C\left( t\right) $ is the length of the current cycle at time $t$, \[ A\left( t\right) /C\left( t\right) \Rightarrow U^{1/\alpha}, \] where $U$ is $U\left( 0,1\right) $. This extends a classical result in renewal theory in the finite mean case which indicates that the limit is $U\left( 0,1\right) $., Comment: 3 pages

Mass-stationarity means that the origin is at a typical location in the mass of a random measure. It is an intrinsic characterisation of Palm versions with respect to stationary random measures. Stationarity is the special case when the random measure is Lebesgue measure. The paper presents constructions of stationary and mass-stationary versions through change of measure and change of origin. Further, the paper considers characterisations of mass-stationarity by distributional invariance under preserving shifts agains stationary independent backgrounds., Comment: To appear in Stochastic Processes and their Applications

We have designed a machine that becomes increasingly better at behaving in underspecified circumstances, in a goal-directed way, on the job, by modeling itself and its environment as experience accumulates. Based on principles of autocatalysis, endogeny, and reflectivity, the work provides an architectural blueprint for constructing systems with high levels of operational autonomy in underspecified circumstances, starting from a small seed. Through value-driven dynamic priority scheduling controlling the parallel execution of a vast number of reasoning threads, the system achieves recursive self-improvement after it leaves the lab, within the boundaries imposed by its designers. A prototype system has been implemented and demonstrated to learn a complex real-world task, real-time multimodal dialogue with humans, by on-line observation. Our work presents solutions to several challenges that must be solved for achieving artificial general intelligence.

Let $B=(B_t)_{t\in {\mathbb{R}}}$ be a two-sided standard Brownian motion. An unbiased shift of $B$ is a random time $T$, which is a measurable function of $B$, such that $(B_{T+t}-B_T)_{t\in {\mathbb{R}}}$ is a Brownian motion independent of $B_T$. We characterise unbiased shifts in terms of allocation rules balancing mixtures of local times of $B$. For any probability distribution $\nu$ on ${\mathbb{R}}$ we construct a stopping time $T\ge0$ with the above properties such that $B_T$ has distribution $\nu$. We also study moment and minimality properties of unbiased shifts. A crucial ingredient of our approach is a new theorem on the existence of allocation rules balancing stationary diffuse random measures on ${\mathbb{R}}$. Another new result is an analogue for diffuse random measures on ${\mathbb{R}}$ of the cycle-stationarity characterisation of Palm versions of stationary simple point processes., Comment: Published in at http://dx.doi.org/10.1214/13-AOP832 the Annals of Probability (http://www.imstat.org/aop/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Let $\xi$ be a random measure on a locally compact second countable topological group and let $X$ be a random element in a measurable space on which the group acts. In the compact case, we give a natural definition of the concept that the origin is a typical location for $X$ in the mass of $\xi$, and prove that when this holds the same is true on sets placed uniformly at random around the origin. This new result motivates an extension of the concept of typicality to the locally compact case where it coincides with the concept of mass-stationarity. We describe recent developments in Palm theory where these ideas play a central role.

We introduce and study invariant (weighted) transport-kernels balancing stationary random measures on a locally compact Abelian group. The first main result is an associated fundamental invariance property of Palm measures, derived from a generalization of Neveu's exchange formula. The second main result is a simple sufficient and necessary criterion for the existence of balancing invariant transport-kernels. We then introduce (in a nonstationary setting) the concept of mass-stationarity with respect to a random measure, formalizing the intuitive idea that the origin is a typical location in the mass. The third main result of the paper is that a measure is a Palm measure if and only if it is mass-stationary., Comment: Published in at http://dx.doi.org/10.1214/08-AOP420 the Annals of Probability (http://www.imstat.org/aop/) by the Institute of Mathematical Statistics (http://www.imstat.org)

We give a deterministic algorithm to construct a graph with no loops (a tree or a forest) whose vertices are the points of a d-dimensional stationary Poisson process S, subset of R^d. The algorithm is independent of the origin of coordinates. We show that (1) the graph has one topological end --that is, from any point there is exactly one infinite self-avoiding path; (2) the graph has a unique connected component if d=2 and d=3 (a tree) and it has infinitely many components if d\ge 4 (a forest); (3) in d=2 and d=3 we construct a bijection between the points of the Poisson process and Z using the preorder-traversal algorithm. To construct the graph we interpret each point in S as a space-time point (x,r)\in\R^{d-1}\times R. Then a (d-1) dimensional random walk in continuous time continuous space starts at site x at time r. The first jump of the walk is to point x', at time r'>r, (x',r')\in S, where r' is the minimal time after r such that |x-x'|<1. All the walks jumping to x' at time r' coalesce with the one starting at (x',r'). Calling (x',r') = \alpha(x,r), the graph has vertex set S and edges {(s,\alpha(s)), s\in S}. This enables us to shift the origin of S^o = S + \delta_0 (the Palm version of S) to another point in such a way that the distribution of S^o does not change (to any point if d = 2 and d = 3; point-stationarity)., Comment: 15 pages. Second version with minor corrections